[Discuss] hard drive reliability

[tmetro-blu at vl.com (Tom Metro)]

Richard Pieri wrote:
> It's like the feedback about Hitachi disks on NewEgg's site.  There
> may be hundreds of reported failure s but that ignores the tens or
> hundreds of thousands of disks that don't fail before their
> warranties expire.

What would be interesting is if NewEgg reported the percentage of drives
that are RMAed. Though even then it would be a limited metric as the RMA
period ends after 30 days (or some similar short period) and the useful
metric is the drive's reliability over its expected lifespan. But this
would at least tell you which drives have DOA problems, infant mortality
problems, and compatibility problems.

I haven't been there in a long time, but I used to use:
http://www.storagereview.com/

for drive reviews and reliability data. They have a reliability survey
feature, but it still will likely have a negative bias due to being
voluntary. If they could somehow fund it, they would need to do a survey
where they randomly select organizations to participate.

Even with the ideal reliability data being available, as someone pointed
out, that doesn't help you if there are variations between batches or
models. (Although with sufficient data, one could likely draw some
predictions based on how consistent a manufacturer has been in the past.
You might have to narrow your criteria to looking at drives using the
same technology and/or produced in the same factories.)


markw at mohawksoft.com wrote:
> If you are saying that a catastrophic failure of a storage device does 
> not reflect poorly upon the manufacturer, I suggest you rethink your 
> position.

A perfectly understandable emotional reaction, but not statistically
significant.

Your argument seems to be predicated on the abrupt nature of the
failure. It presumes that a better manufacture could have designed a
mechanism that would gracefully fail. However, if your particular
failure scenario was never seen in Hitachi's accelerated life tests, are
they still to blame? What if it was seen, but only .01% of the time, so
they decided it wasn't worth adding 2% to the product cost to design
around it. Still Hitachi's fault?

Avoiding failure or failing gracefully can sometimes be possible through
more clever design, but sometimes it just costs more. Other times the
problem is out of the OEM's hands, such as the reliability of a
supplier's component having changed. The manufacturer can be blamed if
they squeezed the costs too much, ignored predictable failures, or
inadequately tested the product. Whether Hitachi has done that can only
be proven through statistics.

Still, perfectly understandable if after seeing a few Hitachi drive
failures that you avoid them for your personal use, but I wouldn't base
a large scale purchase on that data.

 -Tom

-- 
Tom Metro
Venture Logic, Newton, MA, USA
"Enterprise solutions through open source."
Professional Profile: http://tmetro.venturelogic.com/